Skip to main content
  1. My Personal Blog/

A pipeline for creating Geospatial AI datasets

·181 words·
Geospatial Machine Learning
Sambhav Singh Rohatgi
Author
Sambhav Singh Rohatgi
Table of Contents

Motivation
#

I’m working as an industry expert on a research project at UC Irvine, led by Dr. Nadia Ahmed. Our team is studying how satellite imagery can be used to track animal migration. To help my teammates understand the process of geospatial dataset creation, I created a flowchart. This flowchart gives a clear picture of the steps involved in preparing geospatial datasets for Machine Learning.

Explanation and recording
#

I ended up giving an impromptu lecture on it for the team. This included an explanation of my NeurIPS 2023 CCAI tutorial: Automating the creation of LULC datasets for semantic segmentation as well. I have divided the recording into sections for easier viewing. I have put the tutorial recording at the end of the page.

Flowchart
#

Keep in mind that the data used in this flowchart is hypothetical, and some abstractions/assumptions were made to make it easier to understand.

I would recommend viewing it in a separate browser window.

P.S - Miro is awesome!

flowchart
#

flowchart as slides
#

Automating the creation of LULC datasets for semantic segmentation
#

Related

DBSCAN clustering and reverse geocoding for geospatial data
·1184 words
Clustering Visualization Geospatial
Part 2: Deploy your custom MLflow server on GCP with Cloud run and github actions
·1655 words
Mlops Gcp
Part 1: Hosting your custom MLflow server on GCP VM
·1090 words
Mlops Gcp